antibiotic treatment failure in four common infections in ... · pdf filegp’...

Antibiotic treatment failure in four common infectionsin UK primary care 1991-2012: longitudinal analysis

OPEN ACCESS

Craig J Currie professor12, Ellen Berni scientific officer2, Sara Jenkins-Jones research fellow2, ChrisD Poole senior research associate 1, Mario Ouwens principal statistician 3, Stefan Driessen head,biometrics3, Hanka de Voogd head, therapeutic area3, Christopher C Butler professor14, ChristopherLl Morgan epidemiologist 2

1Cochrane Institute of Primary Care and Public Health, Cardiff University, Cardiff, UK; 2Global Epidemiology, Pharmatelligence, Cardiff, UK;3Established Pharmaceuticals, Abbott Healthcare Products, Weesp, Netherlands; 4Nuffield Department of Primary Care Health Sciences, Universityof Oxford, Oxford, UK

AbstractObjective To characterise failure of antibiotic treatment in primary carein the United Kingdom in four common infection classes from 1991 to2012.

Design Longitudinal analysis of failure rates for first line antibioticmonotherapies associated with diagnoses for upper and lower respiratorytract infections, skin and soft tissue infections, and acute otitis media.

SettingRoutine primary care data from the UKClinical Practice ResearchDatalink (CPRD).

Main outcome measures Adjusted rates of treatment failure definedby standardised criteria and indexed to year 1 (1991=100).

Results From 58 million antibiotic prescriptions in CPRD, we analysed10 967 607 monotherapy episodes for the four indications: 4 236 574(38.6%) for upper respiratory tract infections; 3 148 947 (28.7%) forlower respiratory tract infections; 2 568 230 (23.4%) for skin and softtissue infections; and 1 013 856 (9.2%) for acute otitis media. In 1991,the overall failure rate was 13.9% (12.0% for upper respiratory tractinfections; 16.9% for lower respiratory tract infections; 12.8% for skinand soft tissue infections; and 13.9% for acute otitis media). By 2012,the overall failure rate was 15.4%, representing an increase of 12%compared with 1991 (adjusted value indexed to first year (1991) 112,95% confidence interval 112 to 113). The highest rate was seen in lowerrespiratory tract infections (135, 134 to 136). While failure rates werebelow 20% for most commonly prescribed antibiotics (amoxicillin,phenoxymethylpenicillin (penicillin-V), and flucloxacillin), notableincreases were seen for trimethoprim in the treatment of upper respiratorytract infections (from 29.2% in 1991-95 to 70.1% in 2008-12) and forciprofloxacin (from 22.3% in 1991-95 to 30.8% in 2008-12) and cefalexin(from 22.0% in 1991-95 to 30.8% in 2008-12) in the treatment of lower

respiratory tract infections. Failure rates for broad spectrum penicillins,macrolides, and flucloxacillin remained largely stable.

Conclusions From 1991 to 2012, more than one in 10 first line antibioticmonotherapies for the selected infections were associated with treatmentfailure. Overall failure rates increased by 12% over this period, with mostof the increase occurring in more recent years, when antibiotic prescribingin primary care plateaued and then increased.

IntroductionMicrobial resistance to antibiotics has increased alarmingly overrecent decades, prompting the World Health Organization todeclare this a global public health crisis.1 The previous 20 yearshave witnessed a sharp increase in strains resistant to β lactams,2and the same is true of other antimicrobial drugs such as thequinolones and even of antibiotics of last resort such as thecarbapenems. Recent antibiotic use in primary care is the singlemost important risk factor for an infection with a resistantorganism.3 Primary care clinicians, however, seldom reportproblems associated with antibiotic resistance in their ownpractice and regard the problem as often outside their control.4Both primary care clinicians andmembers of the public typicallyregard antibiotic resistance as a problem that largely affectspatients in hospital.4 5

There is a need to better characterise and understand theepidemiology and impact of antibiotic treatment failure—whichmight be representative of antibiotic resistance—over time inthe primary care setting. This is not simple because of the largenumber of antibiotic drugs available; differing infection typesand microbial characteristics; and the difficulty in interpretingadverse outcome signals in population data, which might, over

Correspondence to: C J Currie [email protected]

Extra material supplied by the author (see http://www.bmj.com/content/349/bmj.g5493?tab=related#datasupp)

Appendix: Baseline characteristics for antibiotic monotherapies without identifiable indication [posted as supplied by author]

No commercial reuse: See rights and reprints http://www.bmj.com/permissions Subscribe: http://www.bmj.com/subscribe

BMJ 2014;349:g5493 doi: 10.1136/bmj.g5493 (Published 23 September 2014) of 13

Research

RESEARCH

http://creativecommons.org/licenses/by-nc/3.0/

http://www.bmj.com/content/349/bmj.g5493?tab=related#datasupp

http://www.bmj.com/permissions

http://www.bmj.com/subscribe

http://crossmark.crossref.org/dialog/?doi=10.1136/bmj.g5493&domain=pdf&date_stamp=2014-09-23

time, vary according to changes in host, antibiotic, and pathogenfactors.6 Using a large routine primary care data source, weassessed the failure of first line (initial) antibiotic regimens inthe United Kingdom from 1991 to 2012, alongside an analysisof background antibiotic prescription patterns. The UK is oneof the few countries where there are appropriate data to do so,largely because of the nature of its National Health Service(NHS) and the attendant continuity of data recording. Weexpected to find an increase in treatment failure rates for allcommonly used antibiotics, as previously reported elsewhere.7

One of the motives for this study relates to the macrolideantibiotic clarithromycin, which was introduced internationally25 years ago (1991 in the UK). This anniversary provided anopportunity to assess overall antibiotic practice and effectivenessover time for four common classes of infections for whichclarithromycin is used: upper and lower respiratory tractinfections, skin and soft tissue infections, and acute otitis media.

MethodsData sourceWe used data from the Clinical Practice Research Datalink(CPRD), a longitudinal anonymised research database derivedfrom nearly 700 primary care practices in the UK.8

The database currently contains clinical records from over 14million individuals. The computerised data, recorded in thecourse of routine healthcare by general practitioners andassociated staff, include demographic and lifestyle information,drug prescriptions, medical history, test results, and hospitalreferrals. Diagnoses are recorded with the Read codeclassification, a UK general practice standard, and have beenvalidated in several studies showing a high positive predictivevalue.9 Prescription data are well documented in the database,as they are generated within and automatically recorded by theGP’s practice software. Data in the CPRD, and its predecessorthe General Practice Research Database, are broadlyrepresentative of the UK population in terms of age and sex.10Contributing practices are spread geographically throughout theUK.The CPRD organisation applies data quality markers at patientand general practice levels.11 Patients’ records are deemed tobe of acceptable research quality if they are consistent withregard to age, sex, registration, and event dates and the patienthas been permanently registered with the practice. Contributingpractices are assessed for the completeness, plausibility, andcontinuity of their data and are assigned a date—the“up-to-standard” date—from which their data are judged to beof acceptable quality.The number of clinical records and prescriptions added annuallyto the database increased more than fourfold from 1991 to 2012,while the number of registered patients doubled over the sameperiod. Figure 1 illustrates the overall process of selecting studydata.⇓.

Identification of antibiotic treatmentPrescriptions for antibiotic drugs were selected for the fourinfection classes. From these, we identified episodes ofmonotherapy, defined here as one or more consecutiveprescriptions for a single antibiotic separated by no more than30 days and uninterrupted by prescriptions for other antibioticdrug substances. Of all antibiotic prescriptions, 98% weremonotherapy. Monotherapies were defined as first line if therewere no prescriptions for other antibiotics in the preceding 30days.

Monotherapy episodes were excluded if they were started before1991 or after 2012 or if the interval from the later of the patient’sregistration date or the practice’s up-to-standard date to initiationwas less than 365 days.

Classification of infectionWe selected diagnostic codes relating to the four infectionclasses: upper respiratory tract infections (for example,pharyngitis, laryngitis, tonsillitis, sinusitis), lower respiratorytract infections (for example, pneumonia, bronchitis, whoopingcough), skin and soft tissue infections (for example, cellulitis,impetigo, abscesses), and acute otitis media. Monotherapieswere selected if they had a single associated diagnosis (thenearest to treatment initiation) in one of these classes.The index date was defined as the date of the first prescriptionin the monotherapy episode. More than one episode could beidentified for any individual patient.

Infection rates and antibiotic prescription overtimeIn a separate procedure (not illustrated), we measuredbackground infection rates over time for each infection class asthe proportion of all GP consultations in which the GP recordeda diagnosis of that infection. Additionally, we determined theproportion of these infection related consultations in which anantibiotic was prescribed. This enabled us to assess the patternof prescription over the research period and contextualise thetreatment failure rates.

Study effectiveness endpoint—antibiotictreatment failureAs a proxy measure of effectiveness of treatment and, byimplication, bacterial sensitivity, we evaluated the proportionof antibiotic courses resulting in treatment failure. While thereis no consensus definition of such failure,12 several studies havedefined it as the prescription of a different antibiotic within 30days of the index date.13-15 Other data in the Clinical PracticeResearch Datalink, however, could also be indicative ofunresolved infections. For this investigation, we thereforedefined treatment failure as the earliest occurrence of any of:prescription of a different antibiotic drug within 30 days of thefirst line antibiotic; GP record of admission to hospital with aninfection related diagnosis within 30 days of antibiotic initiation;GP referral to an infection related specialist service within 30days of initiation; GP record of an emergency department visitwithin three days of initiation (the shorter time window beingselected here to increase the probability that the emergencyevent was related to the infection); or GP record of death withan infection related diagnostic code within 30 days of initiation.We did not include reconsultation for an infection within the30 day time window as a signifier of treatment failure becauseof the difficulty, within routine data, of distinguishing betweena continuation of the original infection and the recording of thatinfection as a historic event. Within this primary care settingno information on in vitro microbiological resistance wasavailable as a reason for switching antibiotic regimens.

Statistical analysisFor each year from 1991 to 2012, we determined antibiotictreatment failure rates for the four infection classes and overall.Because the numbers of patients and participating practicesincreased over the research period, we averaged overalltreatment failure rates for each class over the first five yearperiod (1991-95) to create a more stable baseline and compared



RESEARCH



them with the last five year period (2008-12), with χ2 to test thehypothesis of no change in rate.Over the period, there were changes in population age and inthe types of antibiotics used to treat each infection class andsubclass (constituent diagnoses within each class). Thereforewe also report antibiotic treatment failure rates as a standardisedratio (analogous to the standardised mortality rate) of observedto expected treatment failures for each year, indexed to the firstyear (1991=100), with 95% confidence intervals. To calculatethese indexed failure rates, we constructed a matrix of four keystrata (age, sex, antibiotic, and infection subclass) and appliedthe strata specific failure rates observed in 1991 to thecorresponding denominator for each subsequent year andaggregated them to generate the expected number of treatmentfailures. The ratio of observed to expected treatment failureswas then calculated. When necessary, we evaluated theprobability of these changes being due to chance using the χ2test; with this large volume of data, however, even small changeswould be significant at the conventional level.Initial data processing was undertaken with Microsoft SQLServer 2012. Statistical analysis was conducted with IBM SPSSStatistics 20.

ResultsPatients and casesThe Clinical Practice Research Datalink contained records of59 470 173 prescriptions for an antibiotic, issued to 8 146 697patients. We excluded 1 019 488 prescriptions (1.7%) becausethey were not prescribed to patients with research qualityrecords. Of the 331 775 patients (4.1%) accordingly removedfrom the study, 253 903 (76.5%) had records of unacceptableresearch quality because of their temporary registration status.From the antibiotic prescriptions, 10 967 607 first linemonotherapy episodes could be identified with a singleindication within the selected infection classes: 4 236 574 forupper respiratory tract infections (38.6%); 3 148 947 for lowerrespiratory tract infections (28.7%); 2 568 230 for skin and softtissue infections (23.4%); and 1 013 856 for acute otitis media(9.2%) (fig 1).⇓

Baseline characteristicsTable 1⇓ summarises relevant baseline characteristics, groupedby infection class and by early (1991-95) and late (2008-12)time periods. From the early to the late period, the number ofpatients in the study data set increased 2.6-fold, while thenumber of monotherapies increased 2.3-fold. The highestrelative increase in patients and monotherapies was seen forskin and soft tissue infections. Mean age increased by 2 or 3years across all infection classes, and the increase in mean BMIwas about 3. The proportion of current smokers decreased overtime in all groups, as did mean systolic and diastolic bloodpressure.As one might expect, patients treated for lower respiratory tractinfections were the oldest: the mean age was 48.9 in the earlyperiod and 52.1 in the later period. They also formed the leasthealthy group, with the highest mean BMI (25.8 increasing to28.3) and blood pressure (138.4/80.6 mm Hg decreasing to131.7/76.6 mm Hg), and the most current smokers (26.7%decreasing to 23.3%) and users of corticosteroids andbronchodilators (doubling over time).The mean age of patients treated for acute otitis media wassurprisingly old, increasing from 15.6 in the early period to 18.4in the late period (table 1).⇓

Consultation rates for selected infectionclassesGP consultation rates for the four infection classes, with orwithout antibiotic treatment, decreased over time (fig 2, topleft⇓). The all cause consultation rate decreased from 323consultations per 1000 registered patients per year in 1991 to207 in 2012; a decrease of 35.9% (P<0.01). The change inconsultation rates varied by type of infection. Rates ofconsultations per 1000 patients per year decreased from 152 to86 (P<0.01) for upper respiratory tract infection, from 40 to 13(P<0.01) for acute otitis media, and from 80 to 45 (P<0.01) forlower respiratory tract infection. The consultation rate for skinand soft tissue infections increased in a non-linear way from 51to 63 consultations per 1000 patients per year (P<0.01).

Rate of antibiotic prescription by infectionclassFigure 2 (top right) shows the change over time in the proportionof infections in the four selected classes treated with anantibiotic. The overall percentage of consultations in which anantibiotic prescription was issued was 64.3%, with a gradualnon-linear increase from 63.9% in 1991 to 65.6% in 2012,including a fall to 60.8% in 2000. The percentage of infectionstreated with antibiotics differed by infection class, with thegreatest increase being in the smallest class, acute otitismedia—from 63.2% in 1991 to 83.2% in 2012. For upperrespiratory tract infections there was a decrease in the proportiontreated with antibiotics, from 59.0% in 1991 to 55.0% in 2012.

Most commonly prescribed antibioticsThe most commonly prescribed antibiotics were amoxicillin (4650 259; 42.4% of the selected infections), followed byphenoxymethylpenicillin (penicillin-V) (1 289 045; 11.8%; ofwhich 95% were for upper respiratory tract infections) andflucloxacillin (1 214 479; 11.1%; of which 97% were for skinand soft tissue infections).There were clear differences in the use of different antibioticswithin the infection classes (fig 3⇓). Amoxicillin treatmentpredominated in upper and lower respiratory tract infectionsand acute otitis media. In 2012, 81.8% of otitis media infectionstreated with antibiotics received first line amoxicillinmonotherapy (fig 3, bottom right). In 1991, flucloxacillin wasused to treat 28.1% of skin and soft tissue infections treatedwith antibiotics, increasing to 54.2% in 2012 (fig 3, bottom left).For the first and last five years of the study period the set ofmost commonly prescribed antibiotics for each infection classremained the same, although their relative positions changed(table 2⇓). Amoxicillin was by far the most commonlyprescribed throughout.For lower respiratory tract infections, the macrolideerythromycin was the second most commonly prescribedantibiotic in the first five years but was overtaken by anothermacrolide, clarithromycin, in the last five years (table 2).⇓Phenoxymethylpenicillin remained the secondmost commonlyprescribed antibiotic for upper respiratory tract infections in thefirst and last five years. For skin and soft tissue infections andacute otitis media alike, erythromycin was the second mostcommon antibiotic in both periods.

Types of antibiotic treatment failureMost (94.4%) probable antibiotic treatment failures wereidentified by the criterion of switching to an alternative antibioticwithin 30 days of treatment. The next most commonly used



RESEARCH



criterion was referral to a specialist or specialist clinic (4.6% offailures). The proportions of treatment failures identified byeach criterion remained stable over the research period.

Antibiotic treatment failure ratesThe rate of change of treatment failure varied over time andwas non-linear, roughly mirroring the proportion of patientswith an infection who were treated with an antibiotic (fig 2⇓).The overall antibiotic treatment failure rate for the four infectionclasses was 14.7%, with a non-linear increase from 13.9% in1991 to 15.4% in 2012 (fig 2, bottom left). In 1991, the treatmentfailure rates were 12.0% for upper respiratory tract infections;16.9% for lower respiratory tract infections; 12.8% for skin andsoft tissue infections; and 13.9% for acute otitis media. In 2012,treatment failure rates were 12.6% for upper respiratory tractinfections; 21.0% for lower respiratory tract infections; 14.5%for skin and soft tissue infections; and 12.0% for acute otitismedia. For each year, the highest failure rates were seen forlower respiratory tract infections. Failure rates in acute otitismedia and upper respiratory tract infections were comparablefrom 1999 onwards.Within the infection classes, individual types of antibiotic wereassociated with markedly differing failure rates (fig 4⇓, table2⇓). There were some notably high levels of failure; for instance,when used to treat upper respiratory tract infections (a total of56 474 infections) trimethoprim resulted in an overall failurerate of 37.2%, increasing from 24.7% in 1991-95 to 55.9% in2008-12. For lower respiratory tract infections, failure rates forcephalosporins increased markedly—for example, from 22.0% in 1991-95 to 30.8% in 2008-12 for cefalexin—as did thosefor ciprofloxacin (from 22.3% in 1991-95 to 30.8% in 2008-12).Failure rates for flucloxacillin to treat skin and soft tissueinfections did not increase, despite increased use of thisantibiotic. Failure rates for macrolides across the four infectionclasses remained largely stable. In 2012, the antibiotics withthe lowest failure rates for upper respiratory tract infectionswere phenoxymethylpenicillin (9.4%) and amoxicillin (12.2%);for lower respiratory tract infections, these were amoxicillin(18.8%) and clarithromycin (19.2%); for skin and soft tissueinfections, lymecycline (5.8%) and oxytetracycline (7.2%); andfor acute otitis media, amoxicillin (10.3%) and erythromycin(13.4%).

Adjusted rates of failure of antibiotictreatmentAfter adjustment for changes over time in age, sex, infectiondiagnosis, and the antibiotic treatment used, there remainedevidence of an overall increase in antibiotic treatment failurerates (adjusted indexed value of 112, 95% confidence interval112 to 113; fig 2, lower right ⇓). The indexed failure rate hadchanged little by 2000 (99, 98 to 100), the pattern of changebeing more gradual and linear in the period after 2000. As inthe crude failure rates (fig 2, lower left), the indexed failure ratevaried by infection class: 111 (110 to 112) for upper respiratorytract infections; 135 (134 to 136) for lower respiratory tractinfections; 111 (110 to 113) for skin and soft tissue infections;and 96 (94 to 99) for acute otitis media.

DiscussionPrincipal findingsFrom 1991 to 2012, more than one in 10 initial antibioticmonotherapies for four common infections were associated withtreatment failure, as defined by a prespecified set of decision

criteria. For upper and lower respiratory tract infection, skinand soft tissue infection, and acute otitis media, the overallfailure rate increased from 13.9% to 15.4% over this period,and from 14.6% to 15.3% averaged over the earliest and latestfive year periods. With 1991 as the reference year, by 2012there was an adjusted increase in failure rates of 12% for thefour classes overall. Most of the increase was from 2000, whencommunity antibiotic prescribing, which had been falling in thesecond half of the 1990s, plateaued and then once again beganrising.16

The rate of increase in antibiotic treatment failure was lessprominent with the most commonly prescribed antibiotics andin those recommended for first line treatment, such as broadspectrum penicillins (amoxicillin) and the macrolides(clarithromycin and erythromycin). There were notable increasesin failure rates for some antibiotics that are usually notrecommended as first line treatments for the conditions understudy, such as trimethoprim, the cephalosporins, and thequinolones. Such drugs, however, might have been prescribedfor more severely ill and frail patients who had recently beenprescribed a first line drug or from whom resistant organismshad been previously isolated.

Strengths and limitations of the studyThe analysis of antibiotic treatment failure within primary careas a measure of effectiveness has not been attempted before onthis scale. We analysed almost 11 million antibioticmonotherapy treatments for four common infection groups inresearch quality patients.As the Clinical Practice Research Datalink collates data fromroutine practice, there are inevitably missing and erroneous data,coding imperfections, and variations in medical practice. Wemilitated against this to some extent by applying theorganisation’s data quality metrics. Of the patients excludedfrom the study on the basis of the research quality of theirrecords, most (78%) were classed by the Clinical PracticeResearch Datalink as unacceptable on the basis of theirtemporary registration status, which could have resulted in theunder-representation of certain disadvantaged groupswith highermobility and, possibly, greater burdens of illness. The poorquality of data associated with temporary registration, withrespect to patient history and follow-up, however, made thisdecision necessary. The number of antibiotic treatmentsconsequently excluded (1.7%) is unlikely to affect our findings.Exposure to the antibiotic treatments of interest was inferredfrom the prescription records, which could only be consideredas an intention to treat. We could not determine whether thepatient actually redeemed the prescription or whether theantibiotic was taken correctly. The direct effects (as opposed todrug resistance effects) of non-compliance on the failure ratesobserved here would probably be to inflate the number oftreatment failures identified. Any trend towards increasedcompliance, possibly caused by improved patient andpractitioner awareness, might therefore reduce failure rates overtime. We do not know whether increasing use of delayedprescribing influenced our findings.Indications could not be determined by our algorithm for some60% of the monotherapies, largely because the diagnostic codesassociated with these monotherapies described symptoms suchas cough that could not be reliably allocated to an infection classor represented conditions that fell outside our selected infectionclasses. This affected some antibiotics more than others; forexample, 91% of metronidazole monotherapies hadundeterminable indications compared with 45% of



RESEARCH



phenoxymethylpenicillin monotherapies. The sets of diagnosticcodes selected to identify indications were, of necessity, limitedto those in which the site and nature of the infection were clear.We believe that the missing indications should not change theoverall findings as far as the most commonly used antibioticsare concerned. The appendix shows a summary of thephenotypic characteristics of those excluded.The study’s classification of upper respiratory and skin and softtissue infections might have been too broad for meaningfulcomparison, and future analysis would subdivide upperrespiratory tract infection into acute sinusitis andlaryngopharyngitis, the better to differentiate their pathology,andwould distinguish between acne vulgaris—for which chronicrepeat prescriptions are often issued—and other skin and softtissue infections.

Antibiotic treatment failure and bacterialresistanceOur long term data characterising antibiotic treatment failurein primary care, derived from one of the few sources availableto investigate such trends, showed that there was evidence ofan increase in failure rates. Rates could be influenced by severalfactors, including changes in host factors (such as adherence totreatment, the social determinants of health, or reduced immunityassociated with the older profile of the sample in more recentyears), antibiotic related factors (such as changing dose andtreatment duration), and pathogen related factors (such aschanging virulence and antimicrobial resistance).6 The failurerates we observedmight be lower than those observed in hospitalcare because infections in primary care are more often viral orself limiting. Signals of actual bacterial resistance mighttherefore not become apparent against a background ofunnecessary antibiotic use. The possible underestimationunderlines the necessity of making better, diagnosis guided,responsible decisions about antibiotic prescription in dailypractice.Although prescriptions are automatically recorded in the ClinicalPractice Research Datalink with regard to product and quantity,precise dose instructions are not always entered explicitly. Forthis reason, we did not investigate treatment dose, although apreliminary exploration suggested that antibiotic doses mighthave increased over time. This warrants further investigation,with possible revision of the definition of antibiotic treatmentfailure to include dose intensification. When treatment failureincreases despite higher antibiotic doses, this will be of greatconcern.

Implications of studyThis study characterises antibiotic prescribing and treatmentfailure in primary care in the UK. Internationally, however,antibiotic prescribing practices in the community varymarkedly.Antibiotic consumption has beenmost extensively characterisedin the European Union, where, in 2011, this ranged from 11.4defined daily doses per 1000 population in the Netherlands to35.1 defined daily doses in Greece.17 The UK was ranked 17thhighest of the 29 countries surveyed, with prescribing, at 18.8defined daily doses per 1000 population, being more than 60%higher than in the Netherlands. Such variation at a countrylevel—the product of many complex factors, including drugregulatory, educational, and cultural differences—has beencorrelated with resistant invasive infections in hospitals.18 Theresults obtained here for the UK cannot therefore be directlyapplied elsewhere without consideration of local factors. Itremains, however, that the UK is one of few countries that has

the coverage and continuity of community based data to enablea study such as this to be carried out at a patient level.A detailed summary of the current state of antibiotic resistancethroughout the world was published recently by the LancetInfectious Diseases Commission.7While the evidence from thiscommission borders on the alarming, the consensus seems tobe that there is still a chance of dealing with this public healththreat.We have shown that in primary care, where most antibiotics areprescribed, and in a developed country, the impact of increasingantibiotic treatment failure was not as great as we had anticipatedbased on evidence reported from hospital settings with moresevere infections.19 Another interpretation of these data,however, would be that failure rates have indeed increased overthis period and that the observed rates are in reality higherbecause inappropriate or unnecessary prescribing could haveattenuated the failure rates. Given the lack of new antibioticclasses on the horizon, increases in failure rates are troubling.For the recommended first line treatment antibiotics, such asbroad spectrum penicillins and the macrolides, failure rates wererelatively stable over an extended period of more than 20 years,while those for other antibiotics, often not considered first linefor the indications we studied, increased in some notable cases.The highest rates were seen with lower respiratory tractinfections, where affected patients tended to be older and lesshealthy.

ConclusionsWe do not know whether the increases in antibiotic treatmentfailure we identified represent a phenomenon that will resolveor whether this is an early indication of a more dramatic andworrying process. Nevertheless, the finding that treatment failurewas associated with more than one in 10 initial antibiotictreatments in primary care represents a considerable burden onpatients and on the healthcare systems. Not only should ratesof antibiotic resistance continue to be monitored and acted on,trends in failure rates should also be closely scrutinised. Similaranalyses should be carried out for other countries that havevarying rates of antibiotic resistance to further explore theassociation between treatment failure and antibiotic resistance.Our data suggest that primary care physicians could play acentral role in helping to contain rises in antibiotic treatmentfailure by managing patients’ expectations and carefullyconsidering whether each prescription is justified; once thedecision is made to prescribe an antibiotic, the choice shouldfollow current guidelines regarding first line drugs.

We thank Monica S Rocha of Abbott Healthcare Products for her helpwith the figures and in critically reviewing the manuscript.Contributors: CJC, CDP, and MO developed the study protocol. Dataextraction and analysis were carried out by EB and SJJ, supervised byCJC. CDP created the Read code lists. CCB advised on the studyquestion, analysis plan, and interpretation of the study findings, andcontributed to drafting the final report. CLlM provided statistical expertise.Because of the conditions of the data licence, co-authors from thefunding body did not have access to the source data from the ClinicalPractice Research Datalink, though they did have access to processeddata. All other authors had full access to all of the data (includingstatistical reports and tables) in the study. All authors contributed to,read, and approved the final manuscript, and all authors takeresponsibility for the integrity of the data and the accuracy of the dataanalysis. CJC is guarantor.



RESEARCH



What is already known on this topic

Antibiotic resistance is associated with acute treatment failure in hospitals, resulting in considerable excess deaths and healthcare costsPrimary care clinicians and members of the general public typically regard problems associated with antibiotic resistance as hospitalphenomena and largely out of their influence and controlAntibiotic prescribing in primary care is associated with antibiotic resistance at a country level, but the frequency and pattern of failureof treatment in this setting are unknown

What this paper adds

More than one in 10 initial antibiotic monotherapies for upper and lower respiratory tract and skin and soft tissue infections were associatedwith failure over a 22 year period in UK primary careOverall antibiotic treatment failure rates increased from 1991 to 2012 by more than 10%, with most of the increase occurring in morerecent years when antibiotic prescribing plateaued and then increasedThe most commonly prescribed antibiotics were associated with relatively stable failure ratesTreatment failure rates increased in some notable cases, especially when the antibiotic selected was not considered first choice for theindications studied

Funding: The study was funded by Abbott Healthcare Products.Co-authors from the funding body helped design the study andsuggested editorial changes.Competing interests: All authors have completed the Unified CompetingInterest form at www.icmje.org/coi_disclosure.pdf (available on requestfrom the corresponding author) and declare that CDP and CLlM arecontractors of, EB and SJJ are employed by, and CJC is a director ofPharmatelligence, a research consultancy receiving funding from AbbottHealthcare Products for the submitted work and from other healthcarerelated organisations; HDV, MO, and SD are employed by AbbottHealthcare Products; and CCB has received a research grant in kindfrom Alere in support of a publically funded study of COPD and hasreceived honorariums from Alere and the Alliance for the Prudent Useof Antibiotics for presentations on point-of-care testing and diagnosticsin primary care.Ethical approval: Studies using the Clinical Practice Research Datalinkare covered by ethics approval granted by the Trent multicentre researchethics committee (ref 05/MRE04/87). This study was approved by theClinical Practice Research Datalink’s independent scientific advisorycommittee on 31 October 2013, protocol No 13_168R.Data sharing: No additional data available. The study data remain theproperty of the Clinical Practice Research Datalink, provided to theauthors under licence.Transparency: CJC (the manuscript’s guarantor) affirms that themanuscript is an honest, accurate, and transparent account of the studybeing reported; that no important aspects of the study have been omitted;and that any discrepancies from the study as planned (and, if relevant,registered) have been explained.

1 WHO Patient Safety. The evolving threat of antimicrobial resistance: options for action.WHO, 2012.

2 Davies J, Davies D. Origins and evolution of antibiotic resistance. Microbiol Mol Biol Rev2010;74:417-33.

3 Costelloe C, Metcalfe C, Lovering A, Mant D, Hay AD. Effect of antibiotic prescribing inprimary care on antimicrobial resistance in individual patients: systematic review andmeta-analysis. BMJ 2010;340:c2096.

4 Simpson SA, Wood F, Butler CC. General practitioners’ perceptions of antimicrobialresistance: a qualitative study. J Antimicrob Chemother 2007;59:292-29.

5 Hawkings NJ, Wood F, Butler CC. Public attitudes towards bacterial resistance: aqualitative study. J Antimicrob Chemother 2007;59:1155-60.

6 Cosby JL, Francis N, Butler CC. The role of evidence in the decline of antibiotic use forcommon respiratory infections in primary care. Lancet Infect Dis 2007;7:749-56.

7 Laxminarayan R, Duse A, Wattal C, Zaidi AK, Wertheim HF, Sumpradit N, et al. Antibioticresistance—the need for global solutions. Lancet Infect Dis 2013;13:1057-98.

8 Medicines and Healthcare Products Regulatory Agency. Clinical Practice ResearchDatalink. 2014. www.cprd.com/intro.asp.

9 Herrett E, Thomas SL, Schoonen WM, Smeeth L, Hall AJ. Validation and validity ofdiagnoses in the General Practice Research Database: a systematic review. Br J ClinPharmacol 2010;69:4-14.

10 Campbell J, Dedman DJ, Eaton SC, Gallagher AM, Williams TJ. Is the CPRD GOLDpopulation comparable to the U.K. population?Pharmacoepidemiol Drug Saf 2013;22(suppl1):280.

11 Williams T, van Staa T, Puri S, Eaton S. Recent advances in the utility and use of theGeneral Practice Research Database as an example of a UK Primary Care Data resource.Ther Adv Drug Saf 2012;3:89-99.

12 Sánchez García M. Early antibiotic treatment failure. Int J Antimicrob Agents 2009;34(suppl3):S14-9.

13 Longo C, Bartlett G, Macgibbon B, Mayo N, Rosenberg E, Nadeau L, et al. The effect ofobesity on antibiotic treatment failure: a historical cohort study. Pharmacoepidemiol DrugSaf 2013;22:970-6.

14 Hess G, Hill JW, Raut MK, Fisher AC, Mody S, Schein JR, et al. Comparative antibioticfailure rates in the treatment of community-acquired pneumonia: results from a claimsanalysis. Adv Ther 2010;27:743-55.

15 Metsvaht T, Pisarev H, Ilmoja M-L, Parm U, Maipuu L, Merila M, et al. Clinical parameterspredicting failure of empirical antibacterial therapy in early onset neonatal sepsis, identifiedby classification and regression tree analysis. BMC Pediatr 2009;9:72.

16 NHS Prescription Services. Antibiotics national charts 2011. www.nhsbsa.nhs.uk/PrescriptionServices/Documents/PPDPrescribingAnalysisCharts/Antibiotics_National_June_2012.pdf.

17 European Centre for Disease Prevention and Control. Surveillance of antimicrobialconsumption in Europe 2011. ECDC; 2014. www.ecdc.europa.eu/en/publications/Publications/antimicrobial-consumption-europe-surveillance-2011.pdf.

18 Goossens H, Ferech M, Vander Stichele R, Elseviers M; ESAC Project Group. Outpatientantibiotic use in Europe and association with resistance: a cross national database study.Lancet 2005;365:579-87.

19 European Centre for Disease Prevention and Control. Point prevalence survey ofhealthcare associated infections and antimicrobial use in European acute care hospitals2011-2012. ECDC; 2013. www.ecdc.europa.eu/en/publications/publications/healthcare-associated-infections-antimicrobial-use-pps.pdf.

Accepted: 27 August 2014

Cite this as: BMJ 2014;349:g5493This is an Open Access article distributed in accordance with the Creative CommonsAttribution Non Commercial (CC BY-NC 3.0) license, which permits others to distribute,remix, adapt, build upon this work non-commercially, and license their derivative workson different terms, provided the original work is properly cited and the use isnon-commercial. See: http://creativecommons.org/licenses/by-nc/3.0/.



RESEARCH

http://www.icmje.org/coi_disclosure.pdf

http://www.cprd.com/intro.asp

http://www.nhsbsa.nhs.uk/PrescriptionServices/Documents/PPDPrescribingAnalysisCharts/Antibiotics_National_June_2012.pdf



http://www.ecdc.europa.eu/en/publications/Publications/antimicrobial-consumption-europe-surveillance-2011.pdf

http://www.ecdc.europa.eu/en/publications/Publications/antimicrobial-consumption-europe-surveillance-2011.pdf

http://www.ecdc.europa.eu/en/publications/publications/healthcare-associated-infections-antimicrobial-use-pps.pdf

http://www.ecdc.europa.eu/en/publications/publications/healthcare-associated-infections-antimicrobial-use-pps.pdf

http://creativecommons.org/licenses/by-nc/3.0/



Tables

Table 1| Baseline characteristics by class of infection and time period. Figures are numbers (percentage) unless stated otherwise

Acute otitis mediaSkin and soft tissueLower respiratoryUpper respiratory

2008-121991-952008-121991-952008-121991-952008-121991-95

206 413104 275664 580141 121614 701260 517823 143367 000No of patients

260 391157 423919 778197 005907 697430 3331 186 711623 389Antibioticmonotherapies

95 370 (46.2)50 940 (48.9)296 353 (44.6)65 609 (46.5)271 712 (44.2)116 466 (44.7)321 164 (39.0)152 176 (41.5)Male

111 043 (53.8)53 335 (51.1)368 227 (55.4)75 512 (53.5)342 989 (55.8)144 051 (55.3)501 979 (61.0)214 824 (58.5)Female

18.4 (20.5)15.6 (18.6)42.2 (23.6)40.0 (22.8)52.1 (24.3)48.9 (25.3)33.8 (22.8)31.0 (22.0)Mean (SD) age (years)

Smoking:

217 803 (83.7)125 523 (79.7)560 771 (61.0)132 344 (67.2)450 236 (49.6)270 059 (62.8)817 941 (68.9)445 108 (71.4)Never or not known

17 993 (6.9)5 774 (3.7)177 398 (19.3)15 861 (8.0)245 777 (27.1)45 448 (10.5)181 280 (15.3)41 410 (6.6)Ex-smoker

24 595 (9.4)26 126 (16.6)181 609 (19.7)48 800 (24.8)211 684 (23.3)114 826 (26.7)187 490 (15.8)136 871 (22.0)Current

26.2 (7.5)22.8 (6.2)28.8 (6.8)25.8 (5.4)28.3 (6.4)25.8 (5.4)27.7 (6.6)24.9 (5.3)Mean (SD) BMI

Mean (SD) blood pressure (mm Hg):

126.2 (16.8)128.9 (20.3)129.4 (17.0)133.4 (21.9)131.7 (16.8)138.4 (22.0)126.8 (16.6)129.3 (20.7)Systolic

76.2 (10.3)77.8 (11.0)76.3 (10.2)79.0 (11.0)76.6 (10.1)80.6 (10.9)76.3 (10.0)77.7 (10.9)Diastolic

Co-medications:

1 962 (0.7)830 (0.5)15 573 (1.7)2 596 (1.3)134 361 (14.8)26 192 (6.1)27 677 (2.3)5 626 (0.9)Systemiccorticosteroid

12 968 (5.0)7 490 (4.8)58 913 (6.4)7 635 (3.9)264 736 (29.2)89 808 (20.9)102 814 (8.7)33 596 (5.4)Bronchodilator

8 539 (3.3)3 347 (2.1)42 627 (4.6)4 207 (2.1)164 792 (18.2)38 873 (9.0)67 528 (5.7)16 427 (2.6)Inhaled corticosteroid



RESEARCH



Table 2| Average prescription and antibiotic treatment failure rates (%) by infection class and by early and late time period with rank orderin parentheses

Antibiotic treatment failure ratesPrescription rates

2008-121991-952008-121991-95

Upper respiratory tract infections

12.0% (10)11.8% (10)43.6% (1)32.1% (1)Amoxicillin

9.4% (11)10.9% (11)29.3% (2)28.5% (2)Phenoxymethylpenicillin

12.2% (9)12.7% (8)8.0% (3)10.6% (3)Erythromycin

40.5% (2)17.3% (2)1.6% (8)7.8% (4)Others*

14.0% (7)11.9% (9)7.5% (4)6.0% (5)Doxycycline

17.1% (6)15.1% (6)2.7% (6)3.6% (6)Co-amoxiclav

20.4% (3)16.5% (3)1.7% (7)3.4% (7)Cefalexin

19.9% (4)13.9% (7)0.5% (10)3.3% (8)Oxytetracycline

55.9% (1)24.7% (1)0.9% (9)2.1% (9)Trimethoprim

19.1% (5)16.0% (4)0.3% (11)1.9% (10)Cefaclor

13.7% (8)15.1% (5)3.9% (5)0.7% (11)Clarithromycin

Lower respiratory tract infections

17.9% (11)14.9% (11)69.5% (1)49.2% (1)Amoxicillin

60.2% (2)21.7% (4)1.6% (7)12.9% (2)Others†

19.3% (10)17.4% (9)7.1% (3)9.8% (3)Erythromycin

25.2% (7)18.8% (6)4.2% (5)7.9% (4)Co-amoxiclav


30.8% (3)22.0% (3)1.9% (6)4.2% (6)Cefalexin

27.3% (5)21.3% (5)0.5% (11)3.8% (7)Cefaclor

70.1% (1)29.2% (1)0.9% (9)2.7% (8)Trimethoprim

30.8% (4)22.3% (2)1.0% (8)1.8% (9)Ciprofloxacin

24.1% (8)18.0% (8)4.6% (4)1.7% (10)Doxycycline

19.8% (9)18.3% (7)7.9% (2)1.5% (11)Clarithromycin

Skin and soft tissue infections

13.1% (6)13.4% (6)51.8% (1)28.9% (1)Flucloxacillin

13.1% (7)12.3% (7)10.5% (2)14.6% (2)Erythromycin

35.7% (1)21.7% (1)5.1% (5)11.1% (3)Others‡


18.9% (3)16.2% (4)6.6% (4)9.0% (5)Co-amoxiclav

8.9% (9)4.8% (10)1.2% (11)9.0% (6)Minocycline

18.0% (4)14.6% (5)1.3% (10)7.6% (7)Co-fluampicil

26.7% (2)20.2% (2)4.8% (7)6.1% (8)Amoxicillin

12.3% (8)9.0% (8)2.5% (9)2.2% (9)Doxycycline

16.4% (5)18.1% (3)3.4% (8)0.7% (10)Clarithromycin

5.9% (11)0.0% (11)7.6% (3)0.0% (11)Lymecycline

Acute otitis media

10.5% (11)12.1% (11)80.2% (1)59.8% (1)Amoxicillin

14.3% (10)16.0% (9)7.1% (2)10.3% (2)Erythromycin

61.5% (1)19.8% (6)1.0% (6)8.7% (3)Others§

18.1% (8)15.3% (10)4.6% (3)8.6% (4)Co-amoxiclav

21.7% (7)17.9% (8)0.4% (11)3.8% (5)Cefaclor

22.7% (5)20.5% (5)1.3% (5)3.1% (6)Cefalexin

56.0% (2)26.8% (1)0.8% (7)2.7% (7)Trimethoprim

40.6% (3)24.7% (2)0.6% (9)1.7% (8)Phenoxymethylpenicillin

21.7% (6)22.4% (3)0.8% (8)0.6% (9)Ciprofloxacin

15.7% (9)19.7% (7)2.6% (4)0.5% (10)Clarithromycin



RESEARCH



Table 2 (continued)

Antibiotic treatment failure ratesPrescription rates

2008-121991-952008-121991-95

33.2% (4)20.9% (4)0.5% (10)0.3% (11)Flucloxacillin

*64 antibiotics (top 10: ampicillin, ciprofloxacin, flucloxacillin, co-trimethoxazole, azithromycin, cefradine, tetracycline, metronidazole, cefixime, co-trimoxazole).†70 antibiotics (top 10: ampicillin, phenoxymethylpenicillin, cefradine, flucloxacillin, azithromycin, co-trimethoxazole, tetracycline, ofloxacin, cefuroxime, cefixime).‡68 antibiotics (top 10: cefalexin, phenoxymethylpenicillin, trimethoprim, ciprofloxacin, tetracycline, metronidazole, cefaclor, cefradine, nitrofurantoin, clindamycin).§45 antibiotics (top 10: ampicillin, co-trimoxazole, doxycycline, azithromycin, cefixime, co-trimethoxazole, co-fluampicil, oxytetracycline, cefradine, pivampicillin).



RESEARCH



Figures

Fig 1 Data flow diagram for selection of episodes of first line antibiotic monotherapy reported in Clinical Practice ResearchDatalink



RESEARCH



Fig 2 Consultation rates for selected infection classes, proportion of infections in selected classes treated with antibiotic,rate of antibiotic treatment failure by infection class, and fully adjusted and indexed treatment failure rates (indexed to1991=100, antibiotic treatment failure rates adjusted for age, sex, infection subclass, and type of antibiotic treatment used)



RESEARCH



Fig 3 Ten most commonly prescribed antibiotics over observed period, by year and by selected infection



RESEARCH



Fig 4 Rates of failure of antibiotic treatment over observed period by most commonly prescribed antibiotics within eachinfection class



RESEARCH

